♪
You are watching Bloomberg Tech Europe. Coming up, it's Europe's most valuable company and one of the most important cogs in the global AI supply chain. We take a deep dive into ASML, the hidden powerhouse of the AI revolution. We look at how the Dutch company's chipmaking machines underpin major AI breakthroughs, from Nvidia's cutting-edge semiconductors through to ChatGPT. And as the US and China battle it out for AI supremacy, ASML's CEO tells me we are in an AI arms race, and Europe is losing.
If you look at the overall AI ecosystem in Europe, it's very weak anyway you look at it. Today it could be a debate about will the US or China win? It's clear who it isn't. It's Europe. Europe is behind.
This hour we bring you exclusive in-depth interview with Christophe Fouquet, the ASML CEO, at the global headquarters in the Netherlands. Welcome to Bloomberg Tech Europe. Without ASML there would be no ChatGPT, no new iPhone models, and no cutting edge chip innovation. The chipmaking machines it produces are so indispensable to the world tech ecosystem, it is now Europe's most valuable company.
ASML is the company that's at the very center of your phone, your laptop and the AI revolution. And yet most people have never heard of it.
ASML lithography.
ASML began in 1984 in the Netherlands. Four decades later, it's the only company on Earth capable of making the machine that prints the tiniest circuit on the most advanced chips. ASML's earliest ascent was turbocharged by the likes of Intel, Texas Instruments, and AMD, which decided not to do everything in-house. And when it came to building commercially viable systems, a multibillion-dollar bet, that job ultimately went to the newly founded Dutch company. That bet eventually paid off. Today, ASML's crown jewel is the extreme ultraviolet lithography machine. Costing more than $300 million, the size of a double-decker bus, weighing as much as a Boeing 737, and some analysts say containing more than 100,000 parts. The EUV uses microscopic light to etch patterns onto wafers, smaller than a virus, allowing chipmakers to pack billions of transistors onto something the size of a fingernail. ASML sits at the heart of the intricate global chip supply chain. Today's top semiconductor firms, the likes of Nvidia, AMD, and Micron, design the chips. Their designs are sent to foundries like TSMC and Samsung that manufacture the chips, mostly in Asia; and for the most advanced semiconductors, those foundries depend almost entirely on EUV machines from ASML, all made in the Netherlands. The industry's great challenge is the same one it's faced for decades: squeezing more and more computing power into less and less space. That race is summed up in Moore's law: the idea that computing power doubles roughly every two years. With each leap forward getting more expensive and technically challenging, ASML will be crucial to keeping that prediction alive.
When ASML introduced its EUV machines, it helped customers move from 7-nanometer chips down to the three-nanometer chips used by the likes of Nvidia and Apple. Its next big test is whether it can transition to what is called high NA. These newer EUV machines are aimed at pushing chips below two nanometers, giving them even more capabilities, considered crucial for future AI applications. The person leading this effort is Chief Executive Officer Christophe Fouquet. In an exclusive interview recorded on November 14th, Fouquet explained why no other company on the planet can do what his can.
Lithography will always be there because you cannot do anything without lithography. And there will always be the wish to get better lithography. Better lithography means better resolution, better accuracy, and better productivity. If we look at lithography and ASML today, thanks to EUV, we mostly know what to do for our customers for the next 10 to 15 years. And the next 10 to 15 years will still see major innovation in lithography, to continue to work with our customer on cost and transistor density. So that is the first thing. We do low NA today, we just talked about high NA. We may at some point in time introduce even higher numerical aperture tool, Hyper NA, that's the 0.75.
When will Hyper NA come in?
Most probably middle of the next decade. We still have time. But that is still something we are preparing for because, as you said, we have to look long term.
You are also looking into advanced packaging, or 3D packaging, and this is essentially where you get the components, instead of laying them out flat, you build them up a bit like a skyscraper. What is the importance of that? How significant will that be, and why focus on that packaging?
Two reasons why this is becoming important. The first is customers are continuing to drive more transistor density. Moore's law calls for doubling the transistor density; this has been going on for many years and still keeps going. In fact, if you look at AI customers, you mentioned Nvidia before, they want even more than that. So they don't want the transistor density to double every two years; they would like the number of transistors to go up by a factor of 16 every two years. That is what has been happening in the last two or three years. So you're going completely off Moore's law. You need even more transistors. Now, you would still use lithography to try to put as many transistors as possible in an area; that's not enough. And if you cannot get enough transistors doing this, you also have to do that.
On AI, you made a big step by investing almost $1.5 billion in Mistral, one of the leading large language model companies in Europe, certainly a competitor to the likes of OpenAI, Gemini, and Anthropic. You took about an 8-10% stake with that deal. Why?
We saw AI as a huge opportunity for ASML. If you look at ASML, we invest a lot in R&D, $4.5 billion every year to honor. And of course we invest a lot in developing our product. We also invest a lot in software, and we saw that when it comes to development cycle, AI could help us enormously, with the software of course like many companies, but for us to be able to ship our machine to our customer and then maintain those machines, we are writing a huge amount of procedures. We spend a lot of time describing on paper how our tool works so that people can deal with it. And then AI can also help us to really reduce the time our engineers spend on that, so we free them to do other things like integration. So we saw major efficiency opportunities with AI. More importantly, we also believe that AI can improve our product, because our product generates tons of data. Are we in an AI bubble?
When people talk about AI bubble, what do you think?
When people talk about AI bubble, I don't know exactly what they mean. Usually I say there are two ways to look at it. If you look at the industry, I don't think there is a bubble. The impact of AI on the industry is just starting. The impact on the industry will be positive for many, many years to come. So there, there is no bubble. The value of AI, the industrial value of AI, is extremely high and this will be developing over time. There's no bubble because, like I said, we are just starting. Sometimes people talk about a bubble in reference to the stock market, because we have seen some companies getting extremely high valuations as a result of the excitement from AI. There, I think what you will see is more players coming over time. So initially, a few companies were very much the only winners of AI, but because the demand, because the industrial demand will be so high, we need more players. And you will see more and more companies designing chips, AI products, and manufacturing chips. And that could create, of course, some change in the stock market. But the two are very different.
Because that commitment, $300 billion just this year alone, maybe $400 billion next year, that translates into real orders for your kit in the years ahead. You start to see that?
Over time. In the last few months I used to joke with some investors, I say well, I still don't have the equation to translate that into orders for us. It takes a bit of time. You're right, over time demand translates into chip demand for our customer. This translates into a need for more capacity, and this translates into demand for ASML and our peers in the industry.
The infrastructure spend and the kind of deal that OpenAI has done, almost trillion-dollar deals. Just by that alone, which isn't making a profit this year, that makes sense to you?
Investment makes sense overall because you cannot play in AI without investing in hyperscalers, and there may be even a few cycles of investment. Because the investment today is done on certain chips. The chips two or three years from now, if you look at Nvidia, they would be a lot more powerful. And people may be tempted, again in 2027, to invest in hyperscalers to make use of those chips. So you have different cycles. So you have a bit of an arms race, because if you don't invest today, you're out. It's true for a company; I think it is true also for governments. You cannot not play in AI, and I think that's why you have this huge backlog in demand.
